Sugar-sweetened beverage intake, food affordability, and autism spectrum disorder among young U.S. children: A national survey analysis

Problem Considered Evidence on sugar-sweetened beverage (SSB) intake and ASD among young children remains limited, particularly within broader socioeconomic and nutritional contexts. This study examined SSB intake and current ASD among young U.S.children aged 3-5 years and identified population subgroups associated with ASD. Methods We analyzed 2022-2023 National Survey of Children's Health data for 22,118 children aged 3-5 years, representing approximately 10.9 million U.S. children. Survey-weighted regression estimated associations between SSB intake and ASD, while machine-learning analyses identified hierarchical subgroups associated with ASD. Predictive performance was evaluated using held-out validation and repeated cross-validation. Results Weighted ASD prevalence was 2.3% and 31.7% of children lived in household unable to always afford nutritious meals. In the fully adjusted model, the overall SSB association was statistically significant (p=.047), but was sensitive to socioeconomic adjustment and alternative SSB specifications. unable to always afford nutritious meals was associated with higher odds of ASD (aOR:2.09; 95%CI:1.45-3.00). Machine-learning analyses identified substantial subgroup heterogeneity, with the highest observed ASD prevalence among children with weekly (1-6x/week) SSB intake occurring in males older than 4 years unable to always afford nutritious meals (8.0%) and, among those with daily (1-3x/day) intake, in males with low socioeconomic status (8.2%). Discrimination was similar for the machine learning (AUC:0.73, 95%CI:0.69-0.77), regression (AUC:0.74, 95%CI:0.70-0.78), with no significant differences. Conclusion ASD prevalence varied across demographic, socioeconomic, and food-affordability subgroups. Machine learning characterized population heterogeneity and contextualized SSB intake within broader household circumstances. Longitudinal studies and external validation are needed to confirm these patterns.